Skip to content

components.md

Latest commit

 

History

History
52 lines (31 loc) · 1.7 KB

components.md

File metadata and controls

52 lines (31 loc) · 1.7 KB

Components of AI agents

Agents are composed of various components. These include:

  • Agents
  • Tools

Agents

In AutoGen, there are two types of agents:

  • Conversable Agent
  • User Proxy Agent
  • Assistant Agent
  • Group Chat Manager Agent

Conversable Agent

This is the base type of agent available in AutoGen. The core functionality of a Conversable Agent is being able to interact with other agents through messages.

Assistant Agent

An Assistant Agent uses language models such as GPT to solve problems. Using these language models, Assistant Agents can:

  • Write code to solve a problem
  • Evaluate code execution results
  • Suggest corrections and bug fixes.

User Proxy Agent

A User Proxy agent is an agent that interacts with human users. By default, these agents solicit user input at each turn of the conversation. Additionally, User Proxy Agents can:

  • Trigger code execution
  • Generate replies using a language model

Group Chat Manager

A Group Chat Manager is an agent that manages the interactions between various agents. With a Group Chat Manager, you can define the strategy for choosing the next agent to use during a conversation

Tools

Agents can make use of tools to complete tasks.

Some examples of these tools include:

  • Local functions - Enable agents to use code you've already written in your application.
  • Code interpreters - Run code generated by agents to solve a problem.
  • Knowledge-bases - Augment the context and knowledge of your agents by providing it access to data sources. These data sources can include web search engines and databases.

Additional Resources